tlbo-based routing approach for wireless sensor...
TRANSCRIPT
Received: August 22, 2017 94
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
TLBO-Based Routing Approach for Wireless Sensor Networks
Asmae El Ghazi1* Belaîd Ahiod1 Mohammed Abbad1
1LRIT, Associated Unit to CNRST (URAC 29) Faculty of Sciences,
Mohammed V University in Rabat, Morocco
* Corresponding author’s Email: [email protected]
Abstract: Routing in Wireless Sensor Networks (WSN) differs from routing in classical networks, as in WSN links
are unreliable. Indeed, there is no infrastructure and the sensors have limited energy resources and low processing
capacity. Routing problems in many cases are considered NP-hard problems. In these cases, metaheuristics can be
the best solution for this kind of problems. Thus, many metaheuristics are used for routing in WSNs. Ant colony
optimization (ACO) is among the most used metaheuristics, as it showed an excellent performance for routing. Due
to the adaptation difficulties related to the ACO parameters, in this paper we propose a new optimization approach
for routing problem using the Teaching Learning Based Optimization (TLBO). TLBO is a recent nature inspired
metaheuristic based on the influence of the teacher in the class and learners interactions. In this work, TLBO
algorithm is used for routing as it was proposed for continuous optimization problems, despite the fact that the
routing in WSNs is a discrete optimization problem. Our proposed approach is shown to be competitive with ACO-
based approaches in terms of energy consumption and WSN lifetime. The experiments performed in varied scenarios
and the results found, illustrate the efficiency of the TLBO approach against the well-known ACO approaches:
Energy Efficient Ant Based Routing (EEABR) Algorithm and the Improved Ant Colony Optimization Routing
Protocol (IACO).
Keywords: Wireless sensor network, Metaheuristic, Routing, Ant colony optimization, Teaching-learning-based
optimization.
1. Introduction
Wireless Sensor Networks represent an
attractive research area due to several factors such as
the resource-constrained nature of sensor nodes,
interference, energy consumption and a large
number of both commercial and military
applications that this technology offers [1].
Successful network design and deployment include
understanding and modelling several problems
related to these factors, which determine the
accessible extent and data rate of a WSN, as well as
cost and battery lifetime [2]. Therefore, studies,
intended to researchers and graduate fields related to
operations research, applied mathematics and
computer science, give some highlights on a number
of representative network problems in WSN and
focus on their respective optimization problems [3].
NP-hard optimization problems induce the necessity
of optimization methods like metaheuristics. In
WSN, the most studied optimization problem is the
routing problem, as it is a major challenge.
To resolve routing in wireless sensor networks a
diverse range of metaheuristic algorithms are used
including Artificial Bee Colony (ABC) [4], Genetic
Algorithms (GA) [5], Particle Swarm Optimization
(PSO) [6], Harmony Search (HS) [7] and Ant
Colony Optimization (ACO) [8], etc.
The ACO metaheuristic has been successfully
applied to solve routing problems in WSN [9, 10].
Some examples of ant based application include:
Sensor-driven Cost-aware Ant Routing (SC), the
Flooded Forward Ant Routing (FF) algorithm [11],
the Flooded Piggy-backed Ant Routing (FP)
algorithm, the Adaptive ant-based Dynamic Routing
(ADR) [12], the Adaptive Routing (AR) Improved
Adaptive Routing (IAR) algorithm [13], E&D
Received: August 22, 2017 95
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
ANtraTS [14], Energy Efficient Ant Based Routing
(EEABR) Algorithm [15] and the Improved Ant
Colony Optimization Routing Protocol (IACO) [8].
A new efficient metaheuristic has recently been
developed: Teaching Learning Based Optimization
(TLBO) [16, 17]. TLBO algorithm is based on
teacher influence on learners’ output in a class. This
metaheuristic is characterized by the reduced
number of parameters, as the complexity of the
known metaheuristics is the number of parameters
used. In this paper, we propose a new optimization
approach for the routing problems using TLBO as it
was proposed for a continuous optimization
problems. To demonstrate the efficiency of the
proposed TLBO protocol, many simulations have
been performed in MATLAB, using various ranges
of nodes, a different detection zone area, a random
source nodes and a random deployment. The
obtained results demonstrate that our based model
gives a good performance in terms of energy
consumption, data delivery and reliability.
The remaining of the paper is organized as
follows: Section 2 presents the routing approach
based on the ACO. Section 3 describes the main
steps of our routing approach based on the TLBO
and an illustrative example is given in the same
section. Section 4 shows the performance evaluation
of our algorithm. Finally, Section 5 concludes our
work.
2. Routing approach based on the ACO
The authors in [8] have proposed a routing
approach based on ant colony optimization (IACO),
which has provided good results compared with
others ant colony optimization approaches. The ant
approach proposed is based on the following
algorithm of Ant colony optimization, used to find a
good path from the source node to the sink.
2.1 Ant colony optimization algorithm
The ant colony has a high ability to explore and
exploit their environment and using it as a medium
to share information. M. Dorigo and G. Di Caro
have been inspired by ant colonies in nature in order
to create ant colony optimization algorithms [18].
The ACO algorithm basic steps are presented in
Algorithm 1.
An ant moves from a node 𝑠 to a node 𝑟 with the
highest probability value. Every ant walking
between nodes lets an amount of pheromone 𝛿𝜏𝑖𝑗 . This quantity of pheromone changes following
environmental factors (wind, sun...), which are
presented by the evaporation rate parameter 𝜌.
2.2 Routing using ant colony optimization
2.2.1. Improved ant colony optimization
The ACO approach is proposed for routing in
flat networks with static sensors and sink, in order to
minimize the energy consumption. When an event is
detected the information is transmitted to the sink
separated on 𝑁 parts. The ants use the probabilistic
decision rules Eq. (1) in the path from source to the
base station.
𝑃𝑘(𝑟, 𝑠) =
{[𝜏(𝑟,𝑠)]𝛼.[𝜂(𝑟,𝑠)]𝛽.[𝛿(𝑟,𝑠)]𝛾
∑ [𝜏(𝑟,𝑠)]𝛼.[𝜂(𝑟,𝑠)]𝛽.[𝛿(𝑟,𝑠)]𝛾𝑟𝜖𝑅𝑠
𝑖𝑓 𝑘 ∉ 𝑡𝑎𝑏𝑢𝑟
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 (1)
Where 𝜏 (𝑟, 𝑠) is the pheromone value between node
𝑠 and 𝑟, 𝜂 is the first heuristic value related to the
energy of nodes. 𝑅𝑠 are the receiver nodes, the
second heuristic value is 𝛿 used to distinguish the
best neighbor. 𝛼, 𝛽 and 𝛾 are the control parameters
relative to the pheromone influence and the
heuristics information. 𝑡𝑎𝑏𝑢𝑟 is the list of packet
identities already received by node 𝑟. This approach
gives good results, compared to Okdem et al.
approach [9]. However, the novel optimization
technique based on TLBO proposed in this paper
promises better results as demonstrated in the
following sections.
2.2.2. Energy-efficient ant-based routing
Camilo et al. proposed an Energy Efficient Ant
Based Routing (EEABR) Algorithm, which is based
ACO, to improve the energy efficiency of WSNs
Received: August 22, 2017 96
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
and maximize the network lifetime [15]. This
routing algorithm, found the paths between the
source nodes and the sink by forward ants, which
select next hop according to the residual energy of
neighbors and the amount of pheromone left by
backward ant during its journey. Meanwhile, the
number of visited nodes, is introduced in this phase,
which makes those nodes closer to sink node have
more pheromone trail, so that forward ants could
reach the sink node more efficiently. However,
during the early period of transmitting packets, the
forward ants cannot find the optimal or near optimal
paths due to the little difference of the amount of
pheromone trail stored in routing table.
The algorithm uses a good strategy considering
energy levels of the nodes and the lengths of the
routed paths, which have a great contribution to
balancing the energy consumption of nodes and
reducing the time delay. Considering this fact, we
have compared the performance results of our
TLBO approach to the results of the EEABR
algorithm.
3. Our routing approach based on the
TLBO
TLBO approach is a new optimization technique
for routing problem in WSN, proposed to overcome
ant- approach drawbacks such as the parameters
tuning difficulties and the processing time,
considering the sensor limitations. The routing
process begins by the initialization of the population
of paths, then finding the optimum path using TLBO
algorithm described in the next section, then sends
data parts through it.
3.1 The TLBO algorithm
Inspired by the teacher’s influence in the class
and learners interaction, Rao et al. have developed
in 2012 the Teaching learning based optimization
algorithm [17]. This method has outperformed some
of the commonly used metaheuristics regarding
continuous nonlinear numerical optimization
problems. It could be split into two basic parts:
Teacher phase and Learner phase. TLBO has been
used successfully for many multi-objective
problems [19, 20]. Algorithm 2 describes TLBO
process.
TLBO algorithm starts by the initialization of
the population and the termination criterion, and
then researches the best solution using the teacher
and the learner phases. At the end, it verifies if the
termination criterion is satisfied or not.
Algorithm 2 TLBO
begin g ← 0;
initialize_population(P, pop_size)
evaluate(P)
repeat
for i = 1 → pop_size do
{Teacher Phase}
TF = round(1 + r)
Xmean ← calculate_mean_vector(P)
Xteacher ← best_solution(P)
Xnew = Xi + r · (Xteacher − (TF · Xmean))
evaluate(Xnew)
if Xnew better than Xi then
Xi ← Xnew
end if {End of Teacher Phase}
{Learner Phase}
ii ← random(pop_size){ii ≠ i}
if Xi better than Xii then
Xi,new = Xi + r · (Xi − Xii)
else
Xi,new = Xi + r · (Xii − Xi)
end if
evaluate(Xi,new)
if Xi,new better than Xi then
Xi ← Xi,new
end if {End of Learner Phase}
end for
P← remove_duplicate_individuals(P)
g ← g + 1
until (g ≠ num_gen) {termination_condition}
print_best_result(P)
end
3.2 Adapting TLBO for routing in WSN
We implement TLBO for routing problem
following five steps. In the first step, we initialize
the optimization parameters; in the second one we
generate the initial population. The third step
represents the teacher phase and the fourth one
presents the learner phase. Finally, in the last step
after verification, we stop the process or we restart.
Step 1: Initialization of the optimization
parameters considered for routing problems and
definition of the objective function.
The node detecting an event collects information
about available routes that lead to the base station.
By requesting neighbors for a definite time, a
number of paths form the initial population:
𝑃𝑎𝑡ℎ𝑠 = {𝑝1, . . . , 𝑝𝑖 , . . . , 𝑝𝑚}, where every path 𝑝𝑖 is formed a 𝑝𝑖 = {𝑛𝑠, … , 𝑛𝑘, … , 𝑛𝑠𝑖𝑛𝑘}.
Received: August 22, 2017 97
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
Population size (𝑃𝑛): depends on the waiting time
that the source node fixed before requesting
neighbors.
Number of generations (𝐺𝑛): varied depending
on population size (if the population size is small the
algorithm converges faster).
Number of design variables (𝐷𝑛) = 3
Optimization problem: Minimize
𝑓(𝑝𝑖) =∑ 𝐸(𝑝𝑖).∑ 𝑑(𝑝𝑖)
𝑛1
𝑛1
𝑛 (2)
Where 𝑝𝑖 is the corresponding 𝑖𝑡ℎ element in 𝑃𝑛, 𝑛
is length of 𝑝𝑖 , 𝑑 (𝑝𝑖) is approximate delay for 𝑝𝑖, and 𝐸 (𝑝𝑖) is approximate energy consumed by 𝑝𝑖.
Step 2: The energy consumption, number of
hops and delay represent design variables.
According to those ones a random population is
generated, where the number of paths corresponds to
the number of learners. This population is expressed
as:
𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 = (
𝑥1,1 𝑥1,2 𝑥1,3⋮ ⋱ ⋮𝑥𝑚,1 𝑥𝑚,2 𝑥𝑚,3
)
The respective objective values are:
(
𝑓1⋮𝑓𝑖⋮𝑓𝑚)
Step 3 (Teacher phase): The mean of each
population column:
𝑀𝑒𝑎𝑛 = (𝑚1 𝑚2 𝑚3)
Teacher is the best solution for this iteration:
𝑋𝑡𝑒𝑎𝑐ℎ𝑒𝑟 = 𝑚𝑖𝑛 1≤𝑖≤𝑚(𝑓(𝑋𝑖)),
𝑋𝑖 = [𝑥𝑖,1, 𝑥𝑖,2 , 𝑥𝑖,3]
The difference between both values is expressed as:
𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒𝐷 = 𝑟(𝑋𝑡𝑒𝑎𝑐ℎ𝑒𝑟 − 𝑇𝐹𝑀𝑒𝑎𝑛)
According to this difference a new solution is
calculated 𝑋𝑛𝑒𝑤 . 𝑋𝑛𝑒𝑤 is accepted if its objective
value is better than 𝑋𝑜𝑙𝑑.
Figure.1 Initialization of population
Step 4 (Learner phase) Select a random 𝑋𝑗from
population, compare it to the current one 𝑋𝑖, then a
new solution is determined by using Eq. (3).
{𝑋𝑛𝑒𝑤, 𝑖 = 𝑋𝑜𝑙𝑑, 𝑖 + 𝑟𝑖(𝑋𝑖
– 𝑋𝑗), 𝑖𝑓 𝑓(𝑋𝑖) < 𝑓(𝑋𝑗)𝑋𝑛𝑒𝑤, 𝑖 = 𝑋𝑜𝑙𝑑, 𝑖 + 𝑟𝑖(𝑋𝑗
– 𝑋𝑖), 𝑖𝑓 𝑓(𝑋𝑗) < 𝑓(𝑋𝑖)
(3)
Step 5: If the number of generations is achieved
stop process, otherwise go to step 3.
3.3 Illustrative example
In this example, 10 nodes are deployed
randomly. The source node s sends request packets
to neighbors and then initializes the population (see
Fig. 1).
Nodes 1, 2 and 4 receive a request from the
source node. Each one checks the routing table, and
then requests their neighbors and so on until
reaching the sink. After this procedure, a population
of 5 paths is generated. (See Fig. 1)
Population:
Initial population composed by:
𝑝1 = {𝑆, 1, 2, 3, 𝑆𝑖𝑛𝑘} 𝑝2 = {𝑆, 2, 3, 𝑆𝑖𝑛𝑘}
𝑝3 = {𝑆, 4, 5, 𝑆𝑖𝑛𝑘} 𝑝4 = {𝑆, 2, 4, 9, 5, 𝑆𝑖𝑛𝑘} 𝑝5 = {𝑆, 4, 2, 3, 𝑆𝑖𝑛𝑘}
Design Variables:
The considered design variables are:
Consumed Energy: 𝐸(𝑝𝑖), 0 < 𝑖 < 6
Delay: 𝑑 (𝑝𝑖)
Number of hops in 𝑝𝑖
For 𝑝1:
𝐸(𝑝1) = 0.0058
Received: August 22, 2017 98
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
𝑑(𝑝1) = 2.9645
Number of hops in 𝑝1 is 5
After the same calculation for all paths 𝑝𝑖 , the
population is:
𝑃𝑜𝑝 =
(
0.0058 2.9645 50.00460.00470.0071
2.37162.37163.5574
446
0.0060 2.9645 5)
TLBO implementation:
Objective function:
𝑓 =
(
0.00340.00270.00280.00420.0036)
Iteration 1:
{Teacher Phase}
𝑀𝑖𝑛(𝑓) = 0.0027 𝑃𝑜𝑝(2) = [0.0046 2.3716 4]
𝑝2 = {𝑆, 2,3, 𝑆𝑖𝑛𝑘} 𝑋𝑇𝑒𝑎𝑐ℎ𝑒𝑟 = [0.0046 2.3716 4] 𝑀𝑒𝑎𝑛 = (0.0056 2.8459 5.000) 𝑇𝐹 = 1 and 𝑟 = 0.1968
𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒𝐷 = [−0.0007 − 0.3216 − 0.6781]
𝑋𝑛𝑒𝑤,1 = [0.0051 2.6429 5.0000 ] 𝑓(𝑋𝑛𝑒𝑤,1) = 0.0027
𝑓(𝑋𝑜𝑙𝑑,1) = 0.0034 𝑓(𝑋𝑛𝑒𝑤,1) < 𝑓(𝑋𝑜𝑙𝑑,1)
𝑋1 = 𝑋𝑛𝑒𝑤,1 {Learner Phase}
Selecting a random solution 𝑋𝑗 (𝑖 ≠ 𝑗) 𝑗 = 2
𝑋1 = [0.0051 2.6429 5.0000]
𝑋2 = [0.0046 2.3716 4.0000]
𝑓 (𝑋1) = 0.0027
𝑓(𝑋2) = 0.0027
𝑟 = 0.0246
𝑋𝑛𝑒𝑤,1 = 𝑋1 − 𝑟(𝑋2 – 𝑋1) 𝑋𝑛𝑒𝑤,1 = [ 0.0053 2.7404 5.0000]
𝑃𝑜𝑝(1) = 𝑋𝑛𝑒𝑤,1
𝑃𝑜𝑝 =
(
0.0053 2.7404 5.00000.00460.00470.0071
2.37162.37163.5574
4.00004.00006.0000
0.0060 2.9645 5.0000)
Iteration 2:
{Teacher Phase}
𝑀𝑒𝑎𝑛 = ( 0.0055 2.8011 5.0000 )
𝑇𝐹 = 2 and 𝑟 = 0.6799
𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒𝐷 = [−0.0044 − 2.1965 − 4.0794]
𝑋𝑛𝑒𝑤,2 = [0.0002 0.1751 4.0000 ]
𝑓(𝑋𝑛𝑒𝑤,2) = 8.8305𝑒−06
𝑓(𝑋𝑜𝑙𝑑,2) = 0.0027
𝑓(𝑋𝑛𝑒𝑤,2) < 𝑓(𝑋𝑜𝑙𝑑,2)
𝑋2 = 𝑋𝑛𝑒𝑤,2 {Learner Phase}
Selecting a random solution 𝑋𝑗 (𝑖 ≠ 𝑗) 𝑗 = 3
𝑋2 = [ 0.0002 0.1751 4.0000]
𝑋3 = [ 0.0047 2.3716 4.0000]
𝑓(𝑋2) = 8.8305𝑒−06
𝑓(𝑋3) = 0.028
𝑓(𝑋3) > 𝑓(𝑋2) 𝑟 = 0.4299 𝑋𝑛𝑒𝑤,2 = 𝑋2 − 𝑟(𝑋3 – 𝑋2) 𝑋𝑛𝑒𝑤,2 = [−0.0017 − 0.7691 4.0000 ]
𝑃𝑜𝑝(2) = 𝑋𝑛𝑒𝑤,2
𝑃𝑜𝑝 =
(
0.0053 2.7404 5.0000−0.00170.00470.0071
−0.76912.37163.5574
4.00004.00006.0000
0.0060 2.9645 5.0000)
…
After a number of generations:
𝑃𝑜𝑝 =
(
0.0006 0.3834 5.00000.0007−0.0004−0.0007
0.7529−0.0997−0.3933
4.00004.00006.0000
0.0013 0.8347 5.0000)
𝑓 = 1.0𝑒−03 ∗
(
0.04920.12470.01000.04910.2204)
𝑀𝑖𝑛(𝑓) = 0.0100 ∗ 1.0𝑒−03 Best solution is:
𝑃𝑜𝑝(3) = (0.0065 3.2869 5.5438 ) 𝑝3 = {𝑆, 4,5, 𝑆𝑖𝑛𝑘}
Data transmission:
Source node splits data to 𝑁 parts and sent packets
to the sink through the chosen path 𝑝3 as described
in Fig. 2.
If a node of 𝑝3 fails, the previous node in the path
must search a new path, using the same process as
source node.
Received: August 22, 2017 99
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
Figure.2 Data transmission
Figure.3 Warning packets
• Warning Packet:
The sink combines received parts to find
information about detected event. If one or many
parts missing, sink sends a warning packets toward
the source node 𝑠 (See Fig. 3) then s sends back the
missing parts.
4. Experimentations and results
This section evaluates the performance of the
proposed routing algorithm based on the TLBO. We
adopt the same conditions to have good
visualization capabilities for the experimental and
comparison purpose.
4.1 Experimental data
We perform simulations in MATLAB under 64
bits Windows operating system. Experiments are
conducted on a computer with Intel(R) i5 3:20GHz
CPU, and 4GB of RAM. The parameter settings
used in the simulations are shown in Table 1.
The radio energy dissipation model used is
“First Order Radio Model” of Heinzelman et al. [21]
mentioned in Fig. 4.
Table 1. Simulations parameters
Parameter Value
No. of nodes 10, 20, 30, 40, ..., 90,100
Simulation area 200m2, 300m2, … 1000m2
Topology Flat topology
Network type Static network
Deployment type Random deployment
Propagation model Free space radio propagation model
Energy model First Order Radio Model [21]
Initial node energy 10 Joules
No. of simulations 10 runs
Figure 4: First order radio model [21].
The main parameters of the wireless sensor
network model are: the initial energy for each node
is 10𝐽 , energy of electronic transmission 𝐸𝑒𝑙𝑒𝑐 = 50𝑛𝐽/𝑏𝑖𝑡 and amplification 휀𝑎𝑚𝑝 = 100𝑛𝐽/𝑏𝑖𝑡/𝑚2.
In order to monitor a static phenomenon, we
deploy randomly a number of nodes varied
according to the coverage area. A source node
detects an event then, using the routing approaches
transmits the information to the sink. The sink
location is known for the source node (if the sink is
not in the coverage area of the source node).
The experiments performed to evaluate the
TLBO approach proposed and to illustrate its
performance against the ACO approaches [8, 15].
The following section shows the results found.
4.2 Obtained results and comparison
In this section, we present simulation results that
evaluate the proposed approach based on TLBO and
validate its ability of well approximating the energy
consumption in the WSN. In addition two well-
known ACO based WSN routing algorithms, were
used to make comparisons; EEABR and IACO.
The Fig. 5 shows the results found for the
routing approaches ACO, EEABR and TLBO. The
average residual energy increases relatively to the
variation of number of the packets received. It is
Received: August 22, 2017 100
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
(a)
(b)
(c)
(d)
(e)
(f)
Received: August 22, 2017 101
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
(g)
(h)
(i)
Figure.5 Simulation results for different WSNs: (a) 20
nodes, (b) 30 nodes, (c) 40 nodes, (d) 50 nodes, (e) 60
nodes, (f) 70 nodes, (g) 80 nodes, (h) 90 nodes, and (i)
100 nodes
Figure.6 Average residual energy after 256 packets are
received
evident to conclude from the same figure that the
TLBO approach results exceed greatly the IACO
and the EEABR, even when the number of nodes
changes. The strategy of TLBO used to choose the
optimal path, lets the data reach the sink faster by
low energy consumption. Overall, comparing TLBO
and ACO metaheuristics for routing, the TLBO
optimization promises better results.
Fig. 6 shows the average residual energy after
256 packets received by the sink node for different
WSNs having various numbers of nodes. From the
Fig .6, it is seen that the difference in energy levels
increases as the number of nodes in the network
grows.
TLBO algorithm has a basic yet important
advantage, which is the absence of parameters. Thus,
its simplicity makes the adaptation for NP-hard
problems easier than other metaheuristics. Ant
colony optimization is one of the metaheuristics
used for routing problems in WSN and gave good
results. However, the number of parameters in ACO
makes TLBO better suited for routing in WSNs as
illustrate in the Fig. 6.
4.3 Discussions
TLBO is a population-based optimization
technique, which implements a group of solutions to
seek for the optimum solution. Many optimization
methods require algorithm parameters like
population size, number of generations, elite size,
etc. In addition to the common control parameters,
different algorithms require their own algorithm
specific control parameters that affect the
performance of the algorithm. IACO and EEABR
require exponent parameters, pheromone
evaporation rate and reward factor. Unlike other
Received: August 22, 2017 102
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
optimization techniques, TLBO approach proposed
does not require any algorithm parameters to be
tuned, thus making the implementation of TLBO
simpler. TLBO is an algorithm-specific parameter-
less algorithm that uses the best solution of the
iteration to change the current solution in the
population thereby increasing the convergence rate.
TLBO does not divide the population, it uses two
different phases: ‘Teacher Phase’ and ‘Learner
Phase’ and uses the mean value of the population to
update the solution.
TLBO algorithm also has not any special
mechanism to handle the constraints. Considering
this fact, the proposed TLBO approach for routing
in wireless sensor networks is better than the IACO
and the EEABR regarding the algorithms
implementation and adaptation to the routing
problem. In addition as illustrated by the results
found in the Fig. 5 and 6, the TLBO approach
optimize the energy in different WSNs better than
the other approaches.
5. Conclusion
This paper presents a new optimization approach
based on a metaheuristic inspired by teacher and
learners behaviours (TLBO). Teaching learning
based optimization is a recent metaheuristic
proposed for continuous optimization problems.
TLBO is an algorithm-specific parameter-less
algorithm considering this fact, TLBO is efficiently
adapted to the routing in WSNs; despite it is a
discrete optimization problem. The proposed
approach provides good results in terms of energy
consumption and WSN’s lifetime. Various
experimentations were performed using the same
settings, in order to validate the proposal. We
demonstrate that the TLBO approach overcomes
widely the ACO based approaches: IACO and
EEABR, especially in energy consumption, which is
a very crucial factor in the wireless sensors networks.
In the near future, we engage to implement our
algorithm in a real wireless sensor network
environment in order to further investigate the
efficiency of our method.
References
[1] N. Xu, “A survey of sensor network applications”,
IEEE Communications Magazine, Vol.40,
pp.102–114, 2002.
[2] C. Zhao, C. Wu, X. Wang, B.W.K. Ling, K.L.
Teo, J.M. Lee, and K.H. Jung, “Maximizing
lifetime of a wireless sensor network via joint
optimizing sink placement and sensor-to-sink
routing”, Applied Mathematical Modelling, Vol.
49, pp. 319-337,2017.
[3] A. Gogu, D. Nace, A. Dilo, and N. Mertnia,
“Optimization problems in wireless sensor
networks”, In: Proc. of International Conf.
Complex, Intelligent and Software Intensive
Systems (CI-SIS), pp.302– 309, 2011.
[4] P.S. Mann and S. Singh, “Artificial bee colony
metaheuristic for energy-efficient clustering and
routing in wireless sensor networks” Soft
Computing, pp.1-14, 2016.
[5] S. Hussain, A.W. Matin, and O. Islam, “Genetic
algorithm for energy efficient clusters in wireless
sensor networks”, In: Proc. of International Conf.
Information Technology: New Generations,
pp.147–154, 2007.
[6] A. El Ghazi and B. Ahiod, “Particle swarm
optimization compared to ant colony optimization
for routing in wireless sensor networks”, In: Proc.
of International Conf. the Mediterranean
Conference on Information & Communication
Technologies’15, pp.221– 227, 2016.
[7] B. Zeng and Y. Dong, “An improved harmony
search based energy-efficient routing algorithm
for wireless sensor networks” Applied Soft
Computing, pp.135–147, 2016.
[8] A. El Ghazi, B. Ahiod, and A. Ouaarab,
“Improved ant colony optimization routing
protocol for wireless sensor networks”, In: Proc.
of International Conf. Networked Systems,
pp.246–256, 2014.
[9] S. Okdem and D. Karaboga, “Routing in wireless
sensor networks using an ant colony optimization
(aco) router chip”, Sensors, Vol.9, No.2, pp.909–
921, 2009.
[10] K. Fathima and K, Sindhanaiselvan, “Ant colony
optimization based routing in wireless sensor
networks”, International Journal of Advanced
Networking and Applications, Vol.4, No.4, pp.
1686, 2013.
[11] Y. Zhang, L.D. Kuhn, and M.P. Fromherz,
“Improvements on ant routing for sensor
networks”, In: Proc. of International Conf. Ant
Colony Optimization and Swarm Intelligence,
Vol.3172, pp.154–165, 2004.
[12] Y. Lu, G. Zhao, and F. Su, “Adaptive ant-based
dynamic routing algorithm”, In: Proc. of
International Conf. World Congress on
Intelligent Control and Automation (WCICA),
Vol.3, pp. 2694–2697, 2004.
[13] R. Ghasem Aghaei, M.A. Rahman, W. Gueaieb,
and A. El Saddik, “Ant colony-based
reinforcement learning algorithm for routing in
wireless sensor networks”, In: Proc. of
International Conf. Instrumentation and
Received: August 22, 2017 103
International Journal of Intelligent Engineering and Systems, Vol.11, No.1, 2018 DOI: 10.22266/ijies2018.0228.10
Measurement Technology Conference
Proceedings (IMTC), pp.1–6, 2007.
[14] Y.f. Wen, Y.q. Chen, and M. Pan, “Adaptive
ant-based routing in wireless sensor networks
using energy* delay metrics”, Journal of
Zhejiang University-SCIENCE A, Vol.9, No.,
pp.531–538, 2008.
[15] T. Camilo, C. Carreto, J. Silva, and F. Boavida,
“An energy-efficient ant-based routing algorithm
for wireless sensor networks”, Ant Colony
Optimization and Swarm Intelligence, pp. 49–59,
2006.
[16] R. Rao, V. Savsani, and D. Vakharia, “Teaching-
learning-based optimization: A novel method for
constrained mechanical design optimization
problems”, Computer-Aided Design, Vol.43,
No.3, pp.303–315, 2011.
[17] R. Rao, V. Savsani, and D. Vakharia, “Teaching-
learning-based optimization: an optimization
method for continuous non-linear large scale
problems”, Information Sciences, Vol.183, No.1,
pp.1–15, 2012.
[18] M. Dorigo and G. Di Caro, “Ant colony
optimization: a new metaheuristic”, In: Proc. of
International Conf. Congress on Evolutionary
Computation (CEC), Vol.2, pp.1–8, 1999.
[19] V. Toğan and A. Mortazavi, “Sizing
optimization of skeletal structures using
teaching-learning based optimization” An
International Journal of Optimization and
Control: Theories & Applications
(IJOCTA), Vol.7, No.2, pp.130-141, 2017.
[20] B.D Raja, R. Jhala, and V. Patel, “Multi-
objective optimization of a rotary regenerator
using tutorial training and self-learning inspired
teaching-learning based optimization algorithm
(ts-tlbo)”, Applied Thermal Engineering, Vol. 93,
pp. 456–467, 2016.
[21] W.R. Heinzelman, A. Chandrakasan, and H.
Balakrishnan, “Energy-efficient communication
protocol for wireless microsensor networks”, In:
Proc. of International Conf. 33rd Annual Hawaii
International Conference on System Sciences,
pp.1–10, 2000.